Machine Learning: Algorithms & Applications

Visualize univariate distributions using histograms, box plots, and density plots to identify skewness, spread, and modality.

Visualize relationships between variables using scatter plots, pair plots, and correlation heatmaps.

Identify data quality issues such as missing values, outliers, and class imbalance through exploratory analysis.

Identify common patterns of missing data and determine appropriate handling strategies based on data context.

Handle missing data using deletion and simple imputation strategies like mean, median, and mode.

Detect outliers using statistical methods such as the IQR and Z-score.

Handle outliers through removal, capping/winsorization, and log transformation.

Identify and resolve duplicate records and common data inconsistencies in a dataset.

Perform a train-test split and explain its role in preventing data leakage.

Apply feature scaling using normalization (min-max) and standardization (Z-score), and identify when each is appropriate.

Encode categorical variables using one-hot, label, and ordinal encoding, and identify when each is appropriate.

Apply variance thresholding to filter out features with little to no variation.

Use correlation analysis to identify and remove redundant, highly-correlated features.

Apply a complete preprocessing pipeline to a raw real-world dataset, including EDA, cleaning, encoding, scaling, and splitting.

Explain the intuition behind linear regression, including its cost function (MSE), ordinary least squares, and key assumptions.

Explain gradient descent as a general optimization method for fitting machine learning models.

Evaluate regression models using key metrics such as MSE, RMSE, MAE, and R-squared.

Explain k-fold and stratified k-fold cross-validation as robust model evaluation strategies.

Explain the bias-variance tradeoff and how it relates to underfitting and overfitting.

Apply linear regression to a real-world dataset and evaluate its performance.

Interpret the coefficients of a trained linear regression model to explain feature-target relationships.

Explain Ridge (L2) and Lasso (L1) regularization, including how each addresses overfitting and Lasso's feature selection property.

Explain Elastic Net as a combination of Ridge and Lasso and when to prefer each regularization approach.

Apply Ridge and Lasso regression to a dataset and interpret the impact of regularization on model coefficients.

Apply Elastic Net and compare its performance against Ridge, Lasso, and standard linear regression.

Explain the intuition behind logistic regression, including the sigmoid function, log-odds, and decision boundary.

Explain when logistic regression is an appropriate choice for a classification problem.

Define and interpret a confusion matrix and its components (TP, FP, TN, FN).

Compute and interpret accuracy, precision, recall, and F1-score, and explain the precision-recall tradeoff.

Plot and interpret a Receiver Operating Characteristic (ROC) curve and compute the Area Under the Curve (AUC).

Apply logistic regression to a binary classification dataset and evaluate it using a full suite of classification metrics.

Explain strategies for extending logistic regression to multi-class problems, such as one-vs-rest and softmax.

Explain the K-Nearest Neighbors (KNN) algorithm, including distance metrics (Euclidean, Manhattan), the role of 'K', and how predictions are made.

Identify the strengths and weaknesses of KNN, such as the curse of dimensionality, computational cost, and sensitivity to feature scaling.

Apply KNN to a classification dataset, tune the hyperparameter K, and evaluate performance.

Explain the intuition behind Naive Bayes, including Bayes' theorem and the 'naive' conditional independence assumption.

Distinguish between Gaussian, Multinomial, and Bernoulli Naive Bayes and identify appropriate use cases for each.

Apply a Naive Bayes classifier to a text or numerical dataset and evaluate its performance.

Explain the decision tree algorithm, including splitting criteria (Gini impurity, information gain), tree construction, and pruning.

Identify the strengths and weaknesses of decision trees, such as their interpretability and tendency to overfit.

Explain the concept of ensemble learning and the 'bagging' strategy for reducing model variance.

Explain how a Random Forest model evolves from bagged decision trees and why it typically improves generalization.

Explain the core intuition of Support Vector Machines (SVMs): maximizing the margin between classes using a hyperplane and support vectors.

Explain the 'kernel trick' and how it enables SVMs to solve non-linearly separable problems using kernels like RBF and polynomial.

Identify the strengths and weaknesses of SVMs, such as their effectiveness in high-dimensional spaces and sensitivity to feature scaling.

Apply an SVM classifier to a dataset, experiment with different kernels, and evaluate its performance.

Explain K-Means clustering, including centroid initialization, iterative assignment/update steps, and convergence criteria.

Choose the optimal number of clusters (K) for K-Means using the elbow method and silhouette score.

Apply K-Means to a dataset and visualize the resulting clusters.

Explain hierarchical agglomerative clustering and interpret a dendrogram to determine the number of clusters.

Apply hierarchical clustering to a dataset and visualize the results using a dendrogram.

Apply DBSCAN to a dataset and visualize the resulting clusters and identified noise points.

Compare the results of K-Means, hierarchical clustering, and DBSCAN on a single dataset, identifying which to use based on data characteristics.

Explain the 'curse of dimensionality' and the importance of dimensionality reduction for machine learning.

Explain Principal Component Analysis (PCA), including its goal of variance maximization and the concept of principal components.

Apply PCA to a high-dimensional dataset and choose the number of components based on explained variance.

Use PCA as a preprocessing step for a supervised learning model and evaluate its impact on performance and training time.

Perform hyperparameter tuning for a given model using grid search and random search with cross-validation.

Develop a decision-making framework to select an appropriate machine learning algorithm based on problem type and data characteristics.

Execute a full preprocessing and feature engineering pipeline on a new, raw dataset.

Train, evaluate, and compare multiple candidate models on the prepared dataset.

Select and justify the choice of a final model based on performance metrics and problem requirements.